Codex needs additional context references durning reasoning

Codex
1

Codex needs to be able to reference context specific instructions during reasoning and it cannot all come the Agents.md file.

Issue to Fix

During reasoning the model doesn’t understand things like:

  • Model inheritance
  • Utils
  • Helpers
  • Existing sharable code
  • Workflows

What I would like to do is have an easy way to build lists of context specific instructions that point to these listed above. The model could request specifics from my local machine using a smaller model to provide help to the reasoner.

After the prompt response is complete it should pass the finalized prompt and changes to the reviewer. The reviewer should be prompted with specific instructions I have in codex as well for things like styling, code cleanup, and other specific instructions. The reviewer keeps the reasoner in check because all the code I get is massive unmanageable code that I have clean up and reduce down to as much as possible. I am not even kidding I have gotten 200+ lines of code for something that could be done with 5 lines.

Layout

  • Agents.md for the bulk of instructions
  • Context Specific Lists: A list of specific contexts to pull from that I can create for reasoning.
  • The reviewer: a file or comment box that I can give to the reviewer to follow up before completing the task.
2

Have you considered

  • Hooks
  • Skills
  • Plugins
3

Do you have any examples per chance?

i have a few skills that I can run after. I’ve considered writing a plugin but don’t have time atm.

I am not sure what hooks are?

4

Update on Codex Hooks status.

5

Thank you for explaining! Much appreciated!

6

A much better idea that is on my to do list is to implement apropos as needed.

Apropos - used to introduce something that is related to or connected with something that has just been said: (ref)

I use SWI-Prolog daily and there is a very useful but often over looked command named apropos.

The idea is that if the AI knows about a word but needs a bit more technical or specific details, if there were a skill, hook, script etc. or such that given the word, phrase or similar, then more specific details could be returned. The nearest human equivalent would be using man pages, or in simpler cases, the command line option --help

An upside to have apropos incorporated with a programming language is that the result stays current with updates, where as with man pages you have to find the correct version of the page, and for help, you need to have the actual command running. By moving the heavy lifting into a script, you save lots of tokens.

A downside is that it takes work to establish this and maintain it as the output for such would be tailored to an AI not for direct human consumption.